Multimodal embedding models are neural networks that transform multiple modalities, such as text and images, into numerical vectors. They are a crucial building block for semantic search/retrieval systems and retrieval-augmented generation (RAG) and are responsible for the retrieval quality. voyage-multimodal-3 is a state-of-the-art multimodal embedding model that uniquely vectorizes interleaved texts + images while capturing visual features from PDFs, slides, tables, figures, and more, eliminating complex document parsing. It improves retrieval accuracy by an average of 19.63% over the next best-performing multimodal embedding model when evaluated across 3 multimodal retrieval tasks (20 total datasets). Latency is 75 ms for a single query with at most 200 tokens, and throughput is 57M tokens per hour at $0.06 per 1M tokens on an ml.g6.xlarge. Learn more about voyage-multimodal-3 here: https://blog.voyageai.com/2024/11/12/voyage-multimodal-3/